Accessible Math: Best Practices After 25 Years of Research and Development
Total Page:16
File Type:pdf, Size:1020Kb
Accessible Math: Best Practices After 25 Years of Research and Development Steve Noble, University of Louisville Neil Soiffer Talking Cat Software Sam Dooley, Edgar Lozano, Dan Brown Pearson Assessments [email protected], [email protected], [email protected], [email protected], [email protected] Abstract Over the last 25 years, there has been significant progress towards making math accessible. This progress has come from both research projects and software development. We summarize best practices about what has been learned in those years. We also highlight two current accessible math development projects that build upon those practices about speech, braille, and overcoming limited working memory. Keywords STEM Accessibility, Mathematics, MathML, Assistive Technology, Braille Journal on Technology and Persons with Disabilities Santiago, J. (Eds): CSUN Assistive Technology Conference © 2018 California State University, Northridge Accessible Math: Best Practices After 25 Years of Research and Development 285 Introduction Work on making math accessible dates back over 25 years. For references to early work, see Karshmer et.al, 2007. In the last 5-10 years, much of the work has focused on reading and navigating math and the results of that work are embedded in several popular tools including JAWS, NVDA+MathPlayer (Soiffer, 2015), ChromeVox (ChromeVox, 2013), Safari+VoiceOver, TextHELP, and MathJax (Cervone et. al., 2016). With the exception of NVDA+MathPlayer, these systems are limited to working in browsers. MathPlayer+NVDA also works in Word documents and PowerPoint with math created by MathType. In the following section, we cover key topics in math accessibility. The results of research studies that have led to best practices are summarized for each topic. Following this summary, we highlight two current projects that are pushing the state of the art further along to establish best practices in new areas. Discussion Prior work leading to best practice Math accessibility encompasses input techniques, output techniques, and ways of interacting with the mathematical content. These techniques span different modalities such as keyboards, speech, and braille. Research into these techniques and modalities are discussed below. Speech output There are many different ways to speak an expression. Common forms of human speech tend to be ambiguous, relying on the listener either being familiar with the equation or being able to see it. Methods to resolve ambiguity typically involve bracketing words (e.g. “fraction … over … end fraction”), prosody (e.g., ASTOR (Raman, 1994) pauses, rate, and/or pitch changes), Journal on Technology and Persons with Disabilities Santiago, J. (Eds): CSUN Assistive Technology Conference © 2018 California State University, Northridge Accessible Math: Best Practices After 25 Years of Research and Development 286 sounds/spearcons (Murphy, 2010), and/or conventions (e.g., MathPlayer’s SimpleSpeak convention that simple fractions such as x/y speak without bracketing words, but more complex ones will be bracketed). Prosody pauses are particularly important to group subexpressions and produce natural speech. Speed and pitch changes are less important: a ClearSpeak study (Frankel & Brownstein, 2016) found blind and low-vision subjects preferred spoken lexical cues over pitch and rate changes. Bracketing words help resolve ambiguity when you can’t see the expression, but are “verbal noise” to those who can see and may hinder understanding for people with a learning disability (Lewis, Noble & Soiffer, 2010). MathPlayer allows users to specify their disability and will avoid bracketing words for people with a learning disability. In addition to disability, other reasons different speech is needed include terse and verbose forms (experts and learners), subject specific readings (e.g., an x with a line over it would normally be read as “x bar” but if the subject area is known to be a statistics, it would be read as “the mean of x”), and semantic vs syntactic readings (e.g., “x squared” vs. “x superscript 2 baseline” in MathSpeak (Isaacson, Schleppenbach, & Lloyd, 2009)). No studies have been done to evaluate preferences for syntactic and semantic verbal renderings. Braille Output JAWS, NVDA+MathPlayer, and Safari display braille math codes on a refreshable braille display, although none of them indicate the current focus of navigation with the braille math output, as is done with dots 7 and 8 for textual content. While there is no extensive research on the effectiveness of braille input position indicators for math, having the braille change to reflect the current focus of navigation, or on the use of an outline of the math expression in braille, a small user study conducted by Pearson in October 2015 (Pearson, 2015) noted that participants Journal on Technology and Persons with Disabilities Santiago, J. (Eds): CSUN Assistive Technology Conference © 2018 California State University, Northridge Accessible Math: Best Practices After 25 Years of Research and Development 287 were confused when dots 7 and 8 were not used to communicate the user’s input position. The Pearson Accessible Equation Editor (AEE) has since been updated to indicate the current input position using dots 7 and 8, as well as the current editor selection with dot 8 on the braille display to provide tactile feedback similar to what users expect for textual content. Display The use of raster images for math in documents has decreased significantly. On the web, MathML and TeX (often rendered by a polyfill such as MathJax or KaTeX) and SVG is common. These technologies allow the math to scale with the font size so that the math can be enlarged without degrading the quality. Native math in a Google doc and math in Microsoft Word or Apple Pages also scale well. In print, larger math expressions are broken over lines. ASTER supported eliding subexpressions and MathJax has experimented with this technique also. For people with learning disabilities, synchronized highlighting of speech and text has been shown to be helpful. No studies have been done with math, but the MeTRC study (Lewis, Lee, Noble and Garrett, 2013) found that students with learning disabilities made twice as many errors reading math as they did reading text; it seems very likely that synchronized highlighting is important for math also. TextHELP originally used synchronized highlighting with MathPlayer and IE; it now makes use of MathJax to do that. Navigation Navigation of math is now supported by most accessible math software. In the ClearSpeak study (Frankel, Brownstein, & Soiffer 2017), it was the highest rated feature studied. Three forms of navigation were supported: character level, notational (2D structures such as division act like a character when arrowing left/right), and structural (left/right arrow keys move Journal on Technology and Persons with Disabilities Santiago, J. (Eds): CSUN Assistive Technology Conference © 2018 California State University, Northridge Accessible Math: Best Practices After 25 Years of Research and Development 288 from operand to operator to operand). For the latter two modes of navigation, up/down arrows are used to move into and out of 2D notational structures. Participants liked all three and found the different navigation modes useful in different circumstances. Most other systems only support notational navigation. Although not studied, the ability to navigate up and down a column, such as in a matrix, a system of equations, or in an elementary math problem seems very important. When moving around, it is sometimes useful to be able to remember a location. MathPlayer’s navigation supports user-defined placeholders. Although ClearSpeak study participants liked that feature, it was not widely used and was rated lower than most other features. Summaries/Outlines/Overviews of larger expression have been tried a number of times in software. Gillan, et. al., (Gillan, Barraza, Karshmer, & Pazuchanics 2004) did a study that showed providing an outline slowed solution time. Nonetheless, they added outlines to their MathGenie solution because they felt it would be useful. As part of a ClearSpeak navigation study, a summary mode was added to MathPlayer; user feedback from studies was that it was not that useful. The authors feel part of this is because the implementation was crude relative to other features. Input/Editors Entering and editing math employs all of the above techniques and raises additional issues. One issue is that there are many more math characters and notations than there are keys on a keyboard. TeX and ASCIIMath are two input notations for specifying both characters and layout. Because they are linear and use a standard keyboard, they are inherently accessible. However, they only support character-by-character navigation and speech, and typically must be Journal on Technology and Persons with Disabilities Santiago, J. (Eds): CSUN Assistive Technology Conference © 2018 California State University, Northridge Accessible Math: Best Practices After 25 Years of Research and Development 289 separately converted to MathML (which screen readers can speak) and braille to verify they were entered correctly. Until recently, these were the only accessible math editor options. Most direct edit/WYSWYG math editors are not accessible. Two exceptions to this are ChattyInfty (Yamaguchi & Suzuki, 2012) and AEE. Math editors typically use palettes and keyboard shortcuts for all the special characters and notations; for accessibility,